ABSTRACT Cocaine use disorder (CUD) is characterized by the decision to use cocaine at the expense of other activities. Lab-based efforts to address this problem have therefore included cocaine choice self-administration procedures that incorporate a non-drug alternative to model this defining feature. Studies using these procedures have typically scheduled competing reinforcers so that the probabilities are certain. However, such deterministic outcomes are not representative of real-world scenarios in which the consequences from drug-related decisions are often unpredictable. Importantly, decision-making in a dynamic, uncertain context significantly alters the value of choice options and requires continuous updating of option values, which engages learning processes and related corticostriatal networks that might be functioning abnormally in CUD. Decision-making in dynamic environments has been successfully modeled using probabilistic reinforcement-learning choice (PRLC) tasks. The integration of these tasks with reinforcement-learning (RL) modeling has been used to capture moment-to- moment changes in the mechanisms of dynamic choice, and the application of neuroscience techniques has begun to identify the underlying neurobiology. This approach has uncovered biologically-based decision-making abnormalities in multiple brain disorders, but has yet to be systematically applied to the experimental study of CUD, The translation of combined RL and neuroscience approaches to CUD is logical considering the maladaptive choice behavior that typifies the disorder, the varying reinforcement probabilities in cocaine users? natural environments, and the learning impairments and neuroadaptations that have been documented in individuals with CUD. Thus, there are critical gaps in our understanding of the mechanisms underlying dynamic cocaine use decisions, and a strong scientific premise for applying an RL framework to fill these gaps. This project proposes rigorous PRLC tasks, RL modeling, neuromodulation/fMRI neuroimaging techniques and complementary, translational study designs in rats and humans to study dynamic choice in CUD. The first set of cross-species experiments will demonstrate the impact of problematic cocaine use on dynamic decision-making and reveal the neurobehavioral and neurobiological processes underlying this abnormal task performance. The second set of experiments will use a PRLC task in which intravenous cocaine is available as an alternative to the non-drug reinforcer to determine the behavioral and neural ?profile? associated with the decision to use cocaine and reduced cocaine choice during treatment. Amphetamine maintenance and non-drug alternative reinforcer treatments reduce cocaine choice, which will be leveraged here to uncover behavioral and neural mechanisms that can be targeted for future treatment development. This project will have a significant impact on the field by establishing the experimental application of reinforcement-learning theory to the study of maladaptive dynamic decision-making in CUD.